URL: https://doi.org/10.48550/arXiv.2606.23616
Summary
This preprint examines whether GPS-based mobile phone data provide representative estimates of population distribution across 2,478 municipalities in Mexico. It compares coverage bias in a single-platform source, Facebook, and a multi-app aggregator, Veraset, against the 2020 Mexican Population Census. Facebook provides higher and more evenly distributed coverage, while the multi-app data concentrate users in larger, wealthier and more digitally connected municipalities. The analysis also shows that coverage bias is spatially structured and driven by different factors across sources. Using explainable machine learning and spatial statistical models, the study demonstrates that representativeness is both source-specific and spatially dependent, providing evidence to support better bias adjustments in unequal and data-scarce settings.